Was Bernoulli Wrong? On Intuitions about Sample Size
|
|
- Edgar Griffin
- 6 years ago
- Views:
Transcription
1 Journal of Behavioral Decision Making J. Behav. Dec. Making, 13: 133±139 (2000) Was Bernoulli Wrong? On Intuitions about Sample Size PETER SEDLMEIER 1 * and GERD GIGERENZER 2 1 University of Paderborn, Germany 2 Max Planck Institute for Human Development, Berlin, Germany ABSTRACT Recently we proposed an explanation for the apparently inconsistent result that people sometimes take account of sample size and sometimes do not: Human intuitions conform to the `empirical law of large numbers,' which helps to solve what we called `frequency distribution tasks' but not `sampling distribution tasks' (Sedlmeier and Gigerenzer, 1997). Keren and Lewis (2000) do not provide an alternative explanation but present a three-pronged criticism of ours: (1) the normative argument that a larger sample size will not invariably provide more reliable estimates, (2) the descriptive argument that under certain circumstances, people are insensitive to sample size, and (3) the claim that sampling distributions are essential for solving both frequency and sampling distribution tasks. We argue that (1) the normative argument is irrelevant for our descriptive hypothesis and, as a normative claim, only valid for a speci c situation, (2) the descriptive argument is correct but consistent with our review, and (3) is incorrect. Bernoulli's assertion that the intuitions of `even the stupidest man' follow the empirical law of large numbers may have been rather on the optimistic side, but in general the intuitions of the vast majority of people do. Copyright # 2000 John Wiley & Sons, Ltd. KEY WORDS sample size; law of large numbers; statistical reasoning Do people take into account information about sample size in statistical reasoning tasks? The literature exhibits seemingly contradictory results: in one group of studies people do and in another they do not. Recently we suggested an explanation for a substantial part of these inconsistent ndings (Sedlmeier and Gigerenzer, 1997). We proposed the hypothesis that human intuition conforms to the empirical law of large numbers, which expresses the empirically observable fact that larger samples generally (but not always) lead to more accurate estimates of population means (Sedlmeier and Gigerenzer, 1997, p. 35; see also Sedlmeier, 1998, p. 281). We further argued that this intuition works only with what we termed `frequency distribution tasks' and not with `sampling distribution tasks.' Consider, for example, Kahneman and Tversky's (1972) maternity ward task: A certain town is served by two hospitals. In the larger hospital about 45 babies are born each day, and in the smaller hospital about 15 babies are born each day. As you know, about 50% of all babies are boys. The exact percentage of boys, however, varies from day to day. Sometimes it may be higher than 50%, sometimes lower. * Correspondence to: Peter Sedlmeier, University of Paderborn, FB2-Psychology, Paderborn, Germany. CCC 0894±3257/2000/010133±07$17.50 Copyright # 2000 John Wiley & Sons, Ltd.
2 134 Journal of Behavioral Decision Making Vol. 13, Iss. No. 1 In the original, that is, a sampling distribution task, it continues: For a period of 1 year, each hospital recorded the days on which more than 60% of the babies born were boys. Which hospital do you think recorded more such days? The corresponding part in a frequency distribution task reads: Which hospital do you think is more likely to nd on a randomly chosen day that more than 60% of babies born were boys? The relevant di erence between the two tasks lies in the kind of distribution involved. Whereas in the sampling distribution task the relative variance of sampling distributions (of proportions of babies who are boys) has to be judged, in the frequency distribution task two proportions based on frequency distributions (of numbers of babies who are boys and girls) have to be judged. The intuition that larger samples usually lead to more exact estimates of the population mean or proportion (here: 50% boys) helps in the solution of the frequency rather than the sampling distribution tasks. 1 Therefore we derived the hypothesis that researchers who reported that people attend to sample size used predominantly frequency distribution tasks whereas those who concluded that people largely ignore sample size used mainly sampling distribution tasks. This is what we found. The hypothesis that frequency distribution tasks yield higher solution rates than corresponding sampling distribution tasks was subsequently supported in independent studies (Sedlmeier, 1998). Note that the hypothesis that people use an intuition corresponding to the law of large numbers when solving sample size tasks is descriptive, not normative. However, we also addressed the normative argument frequently found in sample size research: that the (mathematical) law of large numbers can be used as a yardstick to evaluate people's judgments. We explained why the law of large numbers cannot serve as such a normative yardstick and pointed out that there exists no general normative model for sample size tasks Ð only partial mathematical justi cations for the superiority of larger samples, such as the variance of the sample mean, Chebychev's inequality, and the central limit theorem. None of these results justify a strictly monotonic relation between sample size and the con dence accruing therefrom. Keren and Lewis now argue that Bernoulli (and we) may have been wrong on the following points: `(i) Normatively, an estimate from a larger sample is not necessarily closer to the corresponding population value that it attempts to estimate, and (ii) Descriptively, under certain circumstances, people are indeed insensitive to sample size' ( p. 125). In addition, they argue that the concept of a sampling distribution is essential for solving frequency distribution tasks. THE NORMATIVE ARGUMENT Keren and Lewis mistakenly think that we argue that the empirical law of large numbers is (1) a normative model that (2) says that an estimate from a larger sample would be invariably closer to the population value. Based on this misunderstanding, they argue against these two points by presenting an example that shows a nonmonotonic relationship between sample size and the accuracy of estimates, using the binomial distribution. From that example, they also draw some implications about statistical power. But (1) the empirical law of large numbers is not a norm Ð we used it as a hypothesis about the intuitions of ordinary people, that is, a descriptive statement. Concerning point (2), Keren and Lewis criticize us (and Bernoulli) that `it is inaccurate to assume that a larger sample size will invariably provide a more reliable estimate than the smaller one' (p. 128; italics ours). This implies incorrectly that we actually made this claim. What we said was `This intuition Ð that larger samples generally lead to more accurate estimates of population means Ð is commonly referred to as the ``empirical law of large numbers'' ' (Sedlmeier and Gigerenzer, 1998, p. 35; italics ours). There is a di erence between generally and invariably. Although Keren and Lewis's normative argument is rather peripheral to our hypothesis about people's intuitions, in the following we clarify it and the related comment on statistical power. 1 To solve the frequency distribution task, the intuition conforming to the empirical law of large numbers would lead to the conclusion that the proportion of boys in the larger hospital can be expected to lie closer to 50% and that therefore, the proportion in the smaller hospital is more likely to have a deviation of 60% or more. Of course, one could apply the intuition repeatedly to solve the sampling distribution task but this implies additional reasoning processes and is apparently not what people commonly do.
3 P. Sedlmeier and G. Gigerenzer Reply 135 Can smaller samples be closer to the population mean than larger ones? There is no normative model that excludes the possibility that a smaller sample may be closer to the population value than a larger sample. The law of large numbers (here Keren and Lewis seem to be in full agreement with us) says nothing about such a comparison; it just says that as the sample size approaches in nity, the probability that the deviation of the mean of the sample from the population mean exceeds an arbitrarily small and xed number approaches 0. As von Mises (1957/1981, p. 115) wrote: ` ``The Law of Large Numbers''... says nothing about the course of a concrete sequence of observations.' Likewise, none of the partial justi cations we mentioned allows one to conclude that as sample size increases, the sample mean monotonically approaches the population mean. How about the empirical law of large numbers? As the name implies, this is an empirical law that everybody can observe, not a normative yardstick. Keren and Lewis have incorrectly understood us as establishing a formal link between the central limit theorem and the empirical law of large numbers, possibly because we spoke of the central limit theorem as a partial mathematical justi cation for the superiority of larger samples. We reiterate that there is no mathematical connection between the two. Exhibit 1 shows a manifestation of the empirical law of large numbers where with increasing sample size the di erence between the sample mean and the population mean generally, but not invariably, decreases ( for similar examples, see Freedman et al., 1991, p. 250, and Freudenthal, 1972, p. 484). Anybody can produce results similar to those shown in Exhibit 1 by tossing a coin 50 times and calculating the proportion of heads after every trial. Of course, one cannot expect that this proportion approaches the population value ( p ˆ 0.5) monotonically. Exhibit 1. Proportions of `successes,' calculated over 50 consecutive results from a Bernoulli distribution (i.e. possible outcomes of 0 and 1) with p ˆ 0.5. For instance, in this special case, the outcomes of the rst four trials were 0, 0, 1, and 1 (with 0 corresponding to `failure' and 1 to `success') yielding proportions of successes of 0, 0, 0.33, and 0.5 for trials 1 to 4, respectively What is the criterion: arbitrarily xed boundaries or mean absolute deviations? Keren and Lewis claim that in the maternity ward task a larger sample does not necessarily provide an estimate that is closer to the population value. In their demonstration, they use a binomial distribution with p ˆ 0.5 and sample sizes from n ˆ 2 to n ˆ 50. In analogy to one speci c version of the maternity ward task, they examine the probability that the outcome of a sampling process lies between 0.4n and 0.6n. They nd that this probability does
4 136 Journal of Behavioral Decision Making Vol. 13, Iss. No. 1 not increase monotonically with increasing sample size. Is this demonstration a good argument against the claim that, generally, the accuracy of an estimate increases with increasing sample size? We think not. Consider sample sizes of n ˆ 2 and n ˆ 3. The possible outcomes for n ˆ 2 are 0, 1, and 2 successes out of 2. How many of these outcomes lie within 0.8 (0.4n) and 1.2 (0.6n) and what is their probability of occurrence? The only outcome that satis es their criterion is 1 and its probability of occurrence is 0.5. Now consider the same for n ˆ 3. Here, none of the possible outcomes lies within 1.2 (0.4n) and 1.8 (0.6n), and therefore, the probability that the result of a sampling process lies within the interval is 0. This is due to the fact that the size of the deviation from the expected mean or proportion of a discrete distribution is expressed in multiples of arbitrarily xed proportions. 2 Moreover, the empirical law of large numbers is about the closeness between an estimate and the corresponding population value and not about a speci c kind of deviation from it. So, we recommend looking at more general measures of deviation. Exhibit 2 shows what happens when one examines just such a measure, the expected values for mean absolute deviations between proportion estimates and corresponding population values. 3 The example illustrates that the nonmonotonicity observed by Keren and Lewis disappears if one does not use arbitrarily xed boundaries. Exhibit 2. Expected values for mean absolute deviations between proportion estimates and corresponding population values for sample sizes from n ˆ 2ton ˆ 50, using a binomial distribution with p ˆ 0.5 (see also footnote 3) 2 In the original maternity ward task, which also deals with a xed deviation from the proportion of 0.5, this problem does not arise, because the size of the larger sample is a multiple of the size of the smaller one (15 versus 45), and the xed boundaries coincide with possible outcomes in the smaller sample ( ˆ 6, and ˆ 9). 3 For any sample size n, the mean absolute deviation (MAD) is calculated as: MAD n ˆ Xn n k pk 1 p n k k n p kˆ0 For instance, to calculate the MAD for n ˆ 2, one has possible values for k of 0, 1, and 2. The absolute deviations of the k/n from a p ˆ 0.5 are 0.5, 0, and 0.5, and their respective probabilities according to the binomial distribution with p ˆ 0.5 are 0.25, 0.5, and Thus MAD 2 ˆ 0:25 0:5 0:5 0 0:25 0:5 ˆ 0:25.
5 P. Sedlmeier and G. Gigerenzer Reply 137 Statistical power Keren and Lewis argue that the nonmonotonicity they found has implications for statistical power. Again, this argument is only indirectly linked to our hypothesis that people's intuitions conform to the empirical law of large numbers. It deals with how formal statistical properties depend on the variation in sample sizes, not with people's intuitions. Here, the issue is the same as in the previous section: the restriction of possible outcome values when expressed in multiples of xed proportions. Consider one of Keren and Lewis's examples in which, using binomial distributions, they nd that for H 0 :m ˆ 0.5 and H 1 :m ˆ 0.7, and alpha ˆ 0.05, the power for n ˆ 18 is larger than that for n ˆ 25. How can this be? The reason is that, because of their discrete values, the possible outcomes closest to and smaller than alpha di er from alpha by di erent amounts. Thus, the power calculation uses di erent `true alphas.' For instance, for n ˆ 18, the value closest to and smaller than alpha is p 13 4 X 4 18 ˆ0:048 with a corresponding power of However, for n ˆ 25, the corresponding value is only half the size: p 18 4 X 4 25 ˆ0:022; and not so surprisingly, the power is not larger than in the prior case Ð If, for n ˆ 25, one took an alpha slightly larger than 0.05, that is, p 17 4 X 4 25 ˆ0:054, then the corresponding power would jump up to Keren and Lewis's argument concerning power shows that one should not slavishly stick to an alpha of 0.05 (or 0.01) because `Good loves the 0.06 nearly as much as the 0.05' (Rosnow and Rosenthal, 1989, p. 1277; see also Gigerenzer, 1993). Anyway, the problem of di erently spaced values in discrete distributions (depending on sample size) ceases to exist with continuous sampling distributions such as the t or the F distribution mostly used in behavioural research. To sum up, the empirical law of large numbers is neither a mathematical theorem nor a norm for how people should solve sample size tasks. Instead, we used it as a hypothesis for the intuitions people use to answer frequency distribution tasks. THE DESCRIPTIVE ARGUMENT Keren and Lewis's descriptive conjecture is quite modest: `under certain circumstances, people are indeed insensitive to sample size' ( p. 125). In our article (Sedlmeier and Gigerenzer, 1997), we started out from exactly this observation Ð that people are sometimes sensitive and sometimes insensitive to sample size Ð and proposed an answer to what `under certain circumstances' means: sampling distribution tasks. The challenging question is not whether or not people are `sometimes' insensitive, but why and when (Gigerenzer, 1996). Our review of the literature indicated that insensitivity is particularly acute when the judgement is about the variability of a sampling distribution, that is, when people were asked to solve sampling distribution tasks. Keren and Lewis mention a study of their own where they used the frequency distribution version of the maternity ward task. Although they do not provide exact values, they report (1) a relatively high proportion of `equally likely' responses and (2) that the participants who did not choose the `equally likely' category generally were sensitive to sample size. Given that they used arbitrarily xed boundaries in their experiment, the choice of `larger sample' was not always normatively correct, but consistent with our hypothesis that people use an intuition conforming to the empirical law of large numbers. The result of their study seems to be within the range of those which we have reviewed (Sedlmeier and Gigerenzer, 1997; Sedlmeier, 1998). In sum, Keren and Lewis's conjecture that people are `sometimes' insensitive to sample size is consistent with our previous review. But the challenging question is, rather, why and when? FREQUENCY DISTRIBUTION TASKS VERSUS SAMPLING DISTRIBUTION TASKS According to Keren and Lewis, both frequency and sampling distribution versions of sample size tasks (e.g. the maternity ward task cited above) require statistical inference for which the concept of sampling distribution is indispensable. Of course, both versions of the task can be solved by recourse to sampling distributions. But they need not. After a fair amount of statistical training, sampling distributions may indeed become a natural way to think about the two kinds of tasks. We agree that the concept of a sampling distribution is necessary to solve the sampling distribution version of the task, but is it also necessary to solve the frequency distribution version? A glance at Exhibit 1 shows that a frequency distribution task can be solved without sampling distributions: from intuitively knowing that generally, the estimate from a larger sample lies closer to the population value than that from a smaller sample. This solution is less elegant than the one relying on the concept of sampling distributions,
6 138 Journal of Behavioral Decision Making Vol. 13, Iss. No. 1 but this seems to be the way laypeople solve the task. It is also evident that the empirically observable law illustrated in Exhibit 1 does not give much help in solving sampling distribution tasks. To sum up, frequency distribution tasks can be solved without recourse to sampling distributions. How? By an intuition that conforms to the empirical law of large numbers, just as Bernoulli proposed. This might be di cult to conceive for statistically trained professionals, whose statistical education has replaced `satis cing' statistical intuitions (Fischbein, 1975). CONCLUSION We have argued that Keren and Lewis's normative argument that `an estimate from a larger sample is not necessarily closer to the corresponding population value' ( p. 125) is of little, if any, relevance to the descriptive hypothesis we have put forward. Nevertheless, we tried to clarify the preconditions of situations in which this can happen. The descriptive argument that `under certain circumstances, people are indeed insensitive to sample size' (p. 125) is entirely consistent with the studies we have reviewed. The argument that `the concept of a sampling distribution is equally essential for answering either of the two versions [i.e. sampling and frequency distribution tasks]' ( p. 128) is incorrect, if `essential' means necessary for solving the task. However, the major issue remains how to explain the discrepant results in research on intuitions on sample size. Keren and Lewis conclude that we are still lacking the theoretical knowledge that enables us to predict under which circumstances people will or will not be sensitive to sample size. We do not think that one has to be so pessimistic. There are several variables that have been shown to in uence sample size judgements to a certain extent (Keren and Lewis, 2000; Sedlmeier and Gigerenzer, 1997) and these variables might also be used in prediction. However, the strongest variable so far is the type of task: one can predict that frequency distribution tasks will be solved more easily than sampling distribution tasks. We have o ered an explanation for why this is so. Human intuition conforms to the empirical law of large numbers, which helps much more in the solution of frequency than sampling distribution tasks. The challenge is now either to nd more evidence for this explanation or to come up with an alternative. Such an alternative does not seem to exist, so far, nor have Keren and Lewis proposed one. Meanwhile, we have elaborated our initial explanation into a computational model based on associative learning mechanisms, the PASS model (Sedlmeier, 1998, 1999). PASS models the cognitive processes underlying human probability judgement for the case of serially encoded events. Consistent with the empirical law of large numbers, PASS's con dence judgements increase with increasing sample size. In a single run, that is, in the process of being presented successive events, PASS's con dence estimates do not always increase strictly monotonically, similar to the example shown in Exhibit 1. The average result over many runs, however, conforms to the results shown in Exhibit 2. Because the input to PASS consists of serially presented events, it can only make judgements about frequency distributions and not sampling distributions. This model provides a possible explanation for why people's intuitions about sample size conform to the empirical law of large numbers. Several hundred years ago, Jacob Bernoulli argued that `even the stupidest man knows by some instinct of nature per se and by no previous instruction' that the greater the number of con rming observations, the surer the conjecture (Gigerenzer et al., 1989, p. 29). Bernoulli may have been rather on the optimistic side, but in general the vast majority of people reason in accord with this intuition. ACKNOWLEDGEMENTS We thank Anita Todd for her comments. Special thanks go again to Berna Eden for her valuable contribution. REFERENCES Fischbein, E. The Intuitive Sources of Probabilistic Thinking in Children, Dordrecht: Reidel, Freedman, D., Pisani, R., Purves, R. and Adhikari, A. Statistics, 2nd edn, New York: Norton, Freudenthal, H. `The ``empirical law of large numbers'' or ``the stability of frequencies'', Educational Studies in Mathematics, 4 (1972), 484±90.
7 P. Sedlmeier and G. Gigerenzer Reply 139 Gigerenzer, G. `The superego, the ego, and the id in statistical reasoning'. In Keren, G. and Lewis, C. (eds), A Handbook for Data Analysis in the Behavioral Sciences: Methodological issues ( pp. 311±339), Hillsdale, NJ: Erlbaum, Gigerenzer, G. `On narrow norms and vague heuristics: A rebuttal to Kahneman and Tversky', Psychological Review, 103 (1996), 592±6. Gigerenzer, G., Swijtink, Z., Porter, T., Daston, L., Beatty, J. and KruÈ ger, L. The Empire of Chance: How probability changed science and everyday life, Cambridge: Cambridge University Press, Kahneman, D. and Tversky, A. `Subjective probability: A judgement of representativeness', Cognitive Psychology, 3 (1972), 430±54. Keren, G. and Lewis, C. `Even Bernoulli might have been wrong: A comment on intuitions about sample size', Journal of Behavioral Decision Making, 13 (2000), 125±132. Rosnow, R. L. and Rosenthal, R. `Statistical procedures and the justi cation of knowledge in psychological science', American Psychologist, 44 (1989), 1276±84. Sedlmeier, P. `The distribution matters: Two types of sample-size tasks', Journal of Behavioral Decision Making, 11 (1998), 281±301. Sedlmeier, P. Improving Statistical Reasoning: Theoretical models and practical implications, Mahwah, NJ: Erlbaum, Sedlmeier, P. and Gigerenzer, G. `Intuitions about sample size: The empirical law of large numbers', Journal of Behavioral Decision Making, 10 (1997), 33±51. von Mises, R. Probability, Statistics and Truth, New York: Dover, (Original work published 1957.) Authors' biographies: Peter Sedlmeier is currently a stand-in professor at the University of MuÈ nster and a Privatdozent at the University of Paderborn. His research includes models of frequency processing, tutorial systems for judgement under uncertainty, and probabilistic models of language processing. He is author of Improving Statistical Reasoning: Theoretical models and practical implications (Erlbaum, 1999). Gerd Gigerenzer is director of the Center for Adaptive Behavior and Cognition at the Max Planck Institute for Human Development. He is the recipient of many prizes, including the AAAS Prize for Behavioral Science Research He studies bounded rationality, ecological rationality, and social rationality. His most recent book is Simple Heuristics that Make us Smart (with P. M. Todd and the ABC Research Group; Oxford University Press, 1999). Authors' addresses: Peter Sedlmeier, University of Paderborn, FB-2 Psychology, Paderborn, Germany. Gerd Gigerenzer, Max Planck Institute for Human Development, Center for Adaptive Behavior and Cognition, Lentzeallee 94, Berlin, Germany.
Intuitions About Sample Size: The Empirical Law of Large Numbers
Journal of Behavioral Decision Making, Vol. 10, 33±51 (1997) Intuitions About Sample Size: The Empirical Law of Large Numbers PETER SEDLMEIER University of Paderborn, Germany GERD GIGERENZER Max Planck
More informationImplicit Information in Directionality of Verbal Probability Expressions
Implicit Information in Directionality of Verbal Probability Expressions Hidehito Honda (hito@ky.hum.titech.ac.jp) Kimihiko Yamagishi (kimihiko@ky.hum.titech.ac.jp) Graduate School of Decision Science
More informationFirst Problem Set: Answers, Discussion and Background
First Problem Set: Answers, Discussion and Background Part I. Intuition Concerning Probability Do these problems individually Answer the following questions based upon your intuitive understanding about
More informationWason's Cards: What is Wrong?
Wason's Cards: What is Wrong? Pei Wang Computer and Information Sciences, Temple University This paper proposes a new interpretation
More informationAn Understanding of Role of Heuristic on Investment Decisions
International Review of Business and Finance ISSN 0976-5891 Volume 9, Number 1 (2017), pp. 57-61 Research India Publications http://www.ripublication.com An Understanding of Role of Heuristic on Investment
More informationReasoning with Uncertainty. Reasoning with Uncertainty. Bayes Rule. Often, we want to reason from observable information to unobservable information
Reasoning with Uncertainty Reasoning with Uncertainty Often, we want to reason from observable information to unobservable information We want to calculate how our prior beliefs change given new available
More informationRecognizing Ambiguity
Recognizing Ambiguity How Lack of Information Scares Us Mark Clements Columbia University I. Abstract In this paper, I will examine two different approaches to an experimental decision problem posed by
More informationEvaluating framing e ects
Journal of Economic Psychology 22 2001) 91±101 www.elsevier.com/locate/joep Evaluating framing e ects James N. Druckman * Department of Political Science, University of Minnesota, 1414 Social Sciences
More informationCognitive Science and Bayes. The main question. Several views. Introduction. The main question Several views Examples. The heuristics view
To Bayes or not to Bayes Radboud University Nijmegen 07-04- 09 The original Bayesian view Are people Bayesian estimators? Or more precisely: should cognitive judgments be viewed as following optimal statistical
More informationOn the diversity principle and local falsifiability
On the diversity principle and local falsifiability Uriel Feige October 22, 2012 1 Introduction This manuscript concerns the methodology of evaluating one particular aspect of TCS (theoretical computer
More informationAmbiguous Data Result in Ambiguous Conclusions: A Reply to Charles T. Tart
Other Methodology Articles Ambiguous Data Result in Ambiguous Conclusions: A Reply to Charles T. Tart J. E. KENNEDY 1 (Original publication and copyright: Journal of the American Society for Psychical
More informationSingle-case probabilities and content-neutral norms: a reply to Gigerenzer
P.B.M. Vranas / Cognition 81 (2001) 105±111 105 COGNITION Cognition 81 (2001) 105±111 www.elsevier.com/locate/cognit Discussion Single-case probabilities and content-neutral norms: a reply to Gigerenzer
More informationDear Participants in the Brunswik Society
Dear Participants in the Brunswik Society As many of you will have noticed, the 2009 meeting has been cancelled. This was done because of my dissatisfaction with the program and the manner in which it
More informationBehavioural Economics University of Oxford Vincent P. Crawford Michaelmas Term 2012
Behavioural Economics University of Oxford Vincent P. Crawford Michaelmas Term 2012 Introduction to Behavioral Economics and Decision Theory (with very large debts to David Laibson and Matthew Rabin) Revised
More informationAn Experimental Comparison of the Fairness Models by Bolton and Ockenfels and by Fehr and Schmidt
An Experimental Comparison of the Fairness Models by Bolton and Ockenfels and by Fehr and Schmidt Dirk Engelmann y and Martin Strobel z Humboldt-Universität zu Berlin January 29, 2000 Abstract In this
More informationWendy J. Reece (INEEL) Leroy J. Matthews (ISU) Linda K. Burggraff (ISU)
INEEL/CON-98-01115 PREPRINT Estimating Production Potentials: Expert Bias in Applied Decision Making Wendy J. Reece (INEEL) Leroy J. Matthews (ISU) Linda K. Burggraff (ISU) October 28, 1998 October 30,
More informationWho Is Rational? Studies of Individual Differences in Reasoning
Book Review/Compte rendu 291 Who Is Rational? Studies of Individual Differences in Reasoning Keith E. Stanovich Mahwah, NJ: Lawrence Erlbaum Associates, 1999. Pp. xvi, 1-296. Hardcover: ISBN 0-8058-2472-3,
More informationCategory Size and Category-Based Induction
Category Size and Category-Based Induction Aidan Feeney & David R. Gardiner Department of Psychology University of Durham, Stockton Campus University Boulevard Stockton-on-Tees, TS17 6BH United Kingdom
More information1 The conceptual underpinnings of statistical power
1 The conceptual underpinnings of statistical power The importance of statistical power As currently practiced in the social and health sciences, inferential statistics rest solidly upon two pillars: statistical
More informationWhen Sample Size Matters: The Influence of Sample Size and Category Variability on Children s and Adults Inductive Reasoning
When Sample Size Matters: The Influence of Sample Size and Category Variability on Children s and Adults Inductive Reasoning Chris A. Lawson (clawson@andrew.cmu.edu) Anna F. Fisher (fisher49@andrew.cmu.edu)
More informationUtility Maximization and Bounds on Human Information Processing
Topics in Cognitive Science (2014) 1 6 Copyright 2014 Cognitive Science Society, Inc. All rights reserved. ISSN:1756-8757 print / 1756-8765 online DOI: 10.1111/tops.12089 Utility Maximization and Bounds
More informationThe Regression-Discontinuity Design
Page 1 of 10 Home» Design» Quasi-Experimental Design» The Regression-Discontinuity Design The regression-discontinuity design. What a terrible name! In everyday language both parts of the term have connotations
More informationBayesian and Frequentist Approaches
Bayesian and Frequentist Approaches G. Jogesh Babu Penn State University http://sites.stat.psu.edu/ babu http://astrostatistics.psu.edu All models are wrong But some are useful George E. P. Box (son-in-law
More informationMath 1680 Class Notes. Chapters: 1, 2, 3, 4, 5, 6
Math 1680 Class Notes Chapters: 1, 2, 3, 4, 5, 6 Chapter 1. Controlled Experiments Salk vaccine field trial: a randomized controlled double-blind design 1. Suppose they gave the vaccine to everybody, and
More informationFrequency versus probability formats in statistical word problems
J.St.B.T. Evans et al. / Cognition 77 (2000) 197±213 197 COGNITION Cognition 77 (2000) 197±213 www.elsevier.com/locate/cognit Frequency versus probability formats in statistical word problems Jonathan
More informationSleeping Beauty is told the following:
Sleeping beauty Sleeping Beauty is told the following: You are going to sleep for three days, during which time you will be woken up either once Now suppose that you are sleeping beauty, and you are woken
More informationAP STATISTICS 2013 SCORING GUIDELINES
AP STATISTICS 2013 SCORING GUIDELINES Question 5 Intent of Question The primary goals of this question were to assess a student s ability to (1) recognize the limited conclusions that can be drawn from
More informationThompson, Valerie A, Ackerman, Rakefet, Sidi, Yael, Ball, Linden, Pennycook, Gordon and Prowse Turner, Jamie A
Article The role of answer fluency and perceptual fluency in the monitoring and control of reasoning: Reply to Alter, Oppenheimer, and Epley Thompson, Valerie A, Ackerman, Rakefet, Sidi, Yael, Ball, Linden,
More informationConfirmation Bias. this entry appeared in pp of in M. Kattan (Ed.), The Encyclopedia of Medical Decision Making.
Confirmation Bias Jonathan D Nelson^ and Craig R M McKenzie + this entry appeared in pp. 167-171 of in M. Kattan (Ed.), The Encyclopedia of Medical Decision Making. London, UK: Sage the full Encyclopedia
More informationTEACHING YOUNG GROWNUPS HOW TO USE BAYESIAN NETWORKS.
TEACHING YOUNG GROWNUPS HOW TO USE BAYESIAN NETWORKS Stefan Krauss 1, Georg Bruckmaier 1 and Laura Martignon 2 1 Institute of Mathematics and Mathematics Education, University of Regensburg, Germany 2
More informationWhat is analytical sociology? And is it the future of sociology?
What is analytical sociology? And is it the future of sociology? Twan Huijsmans Sociology Abstract During the last few decades a new approach in sociology has been developed, analytical sociology (AS).
More informationBottom-Up Model of Strategy Selection
Bottom-Up Model of Strategy Selection Tomasz Smoleń (tsmolen@apple.phils.uj.edu.pl) Jagiellonian University, al. Mickiewicza 3 31-120 Krakow, Poland Szymon Wichary (swichary@swps.edu.pl) Warsaw School
More informationRunning head: INDIVIDUAL DIFFERENCES 1. Why to treat subjects as fixed effects. James S. Adelman. University of Warwick.
Running head: INDIVIDUAL DIFFERENCES 1 Why to treat subjects as fixed effects James S. Adelman University of Warwick Zachary Estes Bocconi University Corresponding Author: James S. Adelman Department of
More informationFAQ: Heuristics, Biases, and Alternatives
Question 1: What is meant by the phrase biases in judgment heuristics? Response: A bias is a predisposition to think or act in a certain way based on past experience or values (Bazerman, 2006). The term
More informationUNIT 4 ALGEBRA II TEMPLATE CREATED BY REGION 1 ESA UNIT 4
UNIT 4 ALGEBRA II TEMPLATE CREATED BY REGION 1 ESA UNIT 4 Algebra II Unit 4 Overview: Inferences and Conclusions from Data In this unit, students see how the visual displays and summary statistics they
More informationQuizzes (and relevant lab exercises): 20% Midterm exams (2): 25% each Final exam: 30%
1 Intro to statistics Continued 2 Grading policy Quizzes (and relevant lab exercises): 20% Midterm exams (2): 25% each Final exam: 30% Cutoffs based on final avgs (A, B, C): 91-100, 82-90, 73-81 3 Numerical
More informationReferences. Christos A. Ioannou 2/37
Prospect Theory References Tversky, A., and D. Kahneman: Judgement under Uncertainty: Heuristics and Biases, Science, 185 (1974), 1124-1131. Tversky, A., and D. Kahneman: Prospect Theory: An Analysis of
More informationThe role of sampling assumptions in generalization with multiple categories
The role of sampling assumptions in generalization with multiple categories Wai Keen Vong (waikeen.vong@adelaide.edu.au) Andrew T. Hendrickson (drew.hendrickson@adelaide.edu.au) Amy Perfors (amy.perfors@adelaide.edu.au)
More informationGeneralization and Theory-Building in Software Engineering Research
Generalization and Theory-Building in Software Engineering Research Magne Jørgensen, Dag Sjøberg Simula Research Laboratory {magne.jorgensen, dagsj}@simula.no Abstract The main purpose of this paper is
More informationRetrospective power analysis using external information 1. Andrew Gelman and John Carlin May 2011
Retrospective power analysis using external information 1 Andrew Gelman and John Carlin 2 11 May 2011 Power is important in choosing between alternative methods of analyzing data and in deciding on an
More information5.2 ESTIMATING PROBABILITIES
5.2 ESTIMATING PROBABILITIES It seems clear that the five-step approach of estimating expected values in Chapter 4 should also work here in Chapter 5 for estimating probabilities. Consider the following
More informationBehavioural models. Marcus Bendtsen Department of Computer and Information Science (IDA) Division for Database and Information Techniques (ADIT)
Behavioural models Cognitive biases Marcus Bendtsen Department of Computer and Information Science (IDA) Division for Database and Information Techniques (ADIT) Judgement under uncertainty Humans are not
More informationLecture 2: Learning and Equilibrium Extensive-Form Games
Lecture 2: Learning and Equilibrium Extensive-Form Games III. Nash Equilibrium in Extensive Form Games IV. Self-Confirming Equilibrium and Passive Learning V. Learning Off-path Play D. Fudenberg Marshall
More informationThe Frequency Hypothesis and Evolutionary Arguments
The Frequency Hypothesis and Evolutionary Arguments Yuichi Amitani November 6, 2008 Abstract Gerd Gigerenzer s views on probabilistic reasoning in humans have come under close scrutiny. Very little attention,
More informationEliminative materialism
Michael Lacewing Eliminative materialism Eliminative materialism (also known as eliminativism) argues that future scientific developments will show that the way we think and talk about the mind is fundamentally
More informationSome Thoughts on the Principle of Revealed Preference 1
Some Thoughts on the Principle of Revealed Preference 1 Ariel Rubinstein School of Economics, Tel Aviv University and Department of Economics, New York University and Yuval Salant Graduate School of Business,
More informationJohn Broome, Weighing Lives Oxford: Oxford University Press, ix + 278pp.
John Broome, Weighing Lives Oxford: Oxford University Press, 2004. ix + 278pp. Weighing Lives is a masterwork that everyone interested in ethics should read. In it, Broome develops a general theory for
More informationUnlike standard economics, BE is (most of the time) not based on rst principles, but on observed behavior of real people.
Behavioral Economics Lecture 1. Introduction and the Methodology of Experimental Economics 1. Introduction Characteristica of Behavioral Economics (BE) Unlike standard economics, BE is (most of the time)
More informationChoose an approach for your research problem
Choose an approach for your research problem This course is about doing empirical research with experiments, so your general approach to research has already been chosen by your professor. It s important
More informationThe Conference That Counts! March, 2018
The Conference That Counts! March, 2018 Statistics, Definitions, & Theories The Audit Process Getting it Wrong Practice & Application Some Numbers You Should Know Objectivity Analysis Interpretation Reflection
More informationPatient Autonomy in Health Care Ethics-A Concept Analysis
Patient Autonomy in Health Care Ethics Patient Autonomy in Health Care Ethics-A Concept Analysis Yusrita Zolkefli 1 1 Lecturer, PAPRSB Institute of Health Sciences, Universiti Brunei Darussalam Abstract
More informationIntroduction to Behavioral Economics Like the subject matter of behavioral economics, this course is divided into two parts:
Economics 142: Behavioral Economics Spring 2008 Vincent Crawford (with very large debts to Colin Camerer of Caltech, David Laibson of Harvard, and especially Botond Koszegi and Matthew Rabin of UC Berkeley)
More informationThe Impact of Continuity Violation on ANOVA and Alternative Methods
Journal of Modern Applied Statistical Methods Volume 12 Issue 2 Article 6 11-1-2013 The Impact of Continuity Violation on ANOVA and Alternative Methods Björn Lantz Chalmers University of Technology, Gothenburg,
More informationComparing Direct and Indirect Measures of Just Rewards: What Have We Learned?
Comparing Direct and Indirect Measures of Just Rewards: What Have We Learned? BARRY MARKOVSKY University of South Carolina KIMMO ERIKSSON Mälardalen University We appreciate the opportunity to comment
More informationTo conclude, a theory of error must be a theory of the interaction between human performance variability and the situational constraints.
The organisers have provided us with a both stimulating and irritating list of questions relating to the topic of the conference: Human Error. My first intention was to try to answer the questions one
More informationField-normalized citation impact indicators and the choice of an appropriate counting method
Field-normalized citation impact indicators and the choice of an appropriate counting method Ludo Waltman and Nees Jan van Eck Centre for Science and Technology Studies, Leiden University, The Netherlands
More informationResponse to the ASA s statement on p-values: context, process, and purpose
Response to the ASA s statement on p-values: context, process, purpose Edward L. Ionides Alexer Giessing Yaacov Ritov Scott E. Page Departments of Complex Systems, Political Science Economics, University
More informationEncoding of Elements and Relations of Object Arrangements by Young Children
Encoding of Elements and Relations of Object Arrangements by Young Children Leslee J. Martin (martin.1103@osu.edu) Department of Psychology & Center for Cognitive Science Ohio State University 216 Lazenby
More informationPower and Effect Size Measures: A Census of Articles Published from in the Journal of Speech, Language, and Hearing Research
Power and Effect Size Measures: A Census of Articles Published from 2009-2012 in the Journal of Speech, Language, and Hearing Research Manish K. Rami, PhD Associate Professor Communication Sciences and
More informationTechnical Specifications
Technical Specifications In order to provide summary information across a set of exercises, all tests must employ some form of scoring models. The most familiar of these scoring models is the one typically
More informationWhat Solution-Focused Coaches Do: An Empirical Test of an Operationalization of Solution-Focused Coach Behaviors
www.solutionfocusedchange.com February, 2012 What Solution-Focused Coaches Do: An Empirical Test of an Operationalization of Solution-Focused Coach Behaviors Coert F. Visser In an attempt to operationalize
More informationThe Role of Causal Models in Statistical Reasoning
The Role of Causal Models in Statistical Reasoning Tevya R. Krynski (tevya@mit.edu) Joshua B. Tenenbaum (jbt@mit.edu) Department of Brain & Cognitive Sciences, Massachusetts Institute of Technology 77
More informationEffects of Sequential Context on Judgments and Decisions in the Prisoner s Dilemma Game
Effects of Sequential Context on Judgments and Decisions in the Prisoner s Dilemma Game Ivaylo Vlaev (ivaylo.vlaev@psy.ox.ac.uk) Department of Experimental Psychology, University of Oxford, Oxford, OX1
More informationCOGNITIVE ILLUSIONS RECONSIDERED
Chapter 109 COGNITIVE ILLUSIONS RECONSIDERED GERD GIGERENZER, RALPH HERTWIG and ULRICH HOFFRAGE Max Planck Institute for Human Development, Berlin, Germany PETER SEDLMEIER University of Chemnitz, Germany
More informationPsy 803: Higher Order Cognitive Processes Fall 2012 Lecture Tu 1:50-4:40 PM, Room: G346 Giltner
Psy 803: Higher Order Cognitive Processes Fall 2012 Lecture Tu 1:50-4:40 PM, Room: G346 Giltner 1 Instructor Instructor: Dr. Timothy J. Pleskac Office: 282A Psychology Building Phone: 517 353 8918 email:
More informationthe examples she used with her arguments were good ones because they lead the reader to the answer concerning the thesis statement.
SAMPLE PAPER 2 Using our PW and CT models, I concluded that Meg Greenfield's essay "In Defense of the Animals" is better than Ron Kline's " A Scientist: I am the enemy". Although both the essays had the
More informationTitle: A new statistical test for trends: establishing the properties of a test for repeated binomial observations on a set of items
Title: A new statistical test for trends: establishing the properties of a test for repeated binomial observations on a set of items Introduction Many studies of therapies with single subjects involve
More informationAdaptive Probability Theory: Human Biases as an
Adaptive Probability Theory: Human Biases as an Adaptation André C. R. Martins Escola de Artes Ciências e Humanidades Universidade de São Paulo Av. Arlindo Bétio, 1000 Parque Ecológico do Tietê - Ermelino
More informationAgenetic disorder serious, perhaps fatal without
ACADEMIA AND CLINIC The First Positive: Computing Positive Predictive Value at the Extremes James E. Smith, PhD; Robert L. Winkler, PhD; and Dennis G. Fryback, PhD Computing the positive predictive value
More informationThe Psychodynamic Approach
WJEC Psychology A-level The Psychodynamic Approach Notes Part 1: An Introduction to The Psychodynamic Approach A01 Introduction and Assumptions: Freud adopted the use of psychic determinism = This is the
More informationHow People Estimate Effect Sizes: The Role of Means and Standard Deviations
How People Estimate Effect Sizes: The Role of Means and Standard Deviations Motoyuki Saito (m-saito@kwansei.ac.jp) Department of Psychological Science, Kwansei Gakuin University Hyogo 662-8501, JAPAN Abstract
More informationImproving statistical estimates used in the courtroom. Precis. Bayes Theorem. Professor Norman Fenton. Queen Mary University of London and Agena Ltd
Improving statistical estimates used in the courtroom Professor Norman Fenton Queen Mary University of London and Agena Ltd Address: Queen Mary University of London School of Electronic Engineering and
More informationChapter 02 Developing and Evaluating Theories of Behavior
Chapter 02 Developing and Evaluating Theories of Behavior Multiple Choice Questions 1. A theory is a(n): A. plausible or scientifically acceptable, well-substantiated explanation of some aspect of the
More informationThe secular increase in test scores is a ``Jensen e ect''
Personality and Individual Differences 30 (2001) 553±559 www.elsevier.com/locate/paid The secular increase in test scores is a ``Jensen e ect'' Roberto Colom *, Manuel Juan-Espinosa, LuõÂ s F. GarcõÂ a
More informationAsch Model Answers. Aims and Context
Asch Model Answers Aims and Context Conformity is yielding to group pressure to fit in with others or come to a decision on what to do if unsure based on other people s actions. One of the earliest studies
More informationCan Ordinal Utility Exist?
Can Ordinal Utility Exist? S. Kemp 1, and R. C. Grace 1 1 Psychology Department, University of Canterbury, Christchurch, New Zealand. Keywords: Ordinal utility, measurement, sorting procedures. EXTENDED
More informationExamples of Feedback Comments: How to use them to improve your report writing. Example 1: Compare and contrast
Examples of Feedback Comments: How to use them to improve your report writing This document contains 4 examples of writing and feedback comments from Level 2A lab reports, and 4 steps to help you apply
More informationMath HL Chapter 12 Probability
Math HL Chapter 12 Probability Name: Read the notes and fill in any blanks. Work through the ALL of the examples. Self-Check your own progress by rating where you are. # Learning Targets Lesson I have
More information[1] provides a philosophical introduction to the subject. Simon [21] discusses numerous topics in economics; see [2] for a broad economic survey.
Draft of an article to appear in The MIT Encyclopedia of the Cognitive Sciences (Rob Wilson and Frank Kiel, editors), Cambridge, Massachusetts: MIT Press, 1997. Copyright c 1997 Jon Doyle. All rights reserved
More informationOn Narrow Norms and Vague Heuristics: A Reply to Kahneman and Tversky (1996)
Published in: Psychological Review, 103 (3), 1996, 592 596. www.apa.org/journals/rev/ 1996 by the American Psychological Association, 0033-295X. On Narrow Norms and Vague Heuristics: A Reply to Kahneman
More informationMARK SCHEME for the May/June 2011 question paper for the guidance of teachers 9699 SOCIOLOGY
UNIVERSITY OF CAMBRIDGE INTERNATIONAL EXAMINATIONS GCE Advanced Subsidiary Level and GCE Advanced Level MARK SCHEME for the May/June 2011 question paper for the guidance of teachers 9699 SOCIOLOGY 9699/23
More informationThe Ontogeny and Durability of True and False Memories: A Fuzzy Trace Account
JOURNAL OF EXPERIMENTAL CHILD PSYCHOLOGY 71, 165 169 (1998) ARTICLE NO. CH982468 The Ontogeny and Durability of True and False Memories: A Fuzzy Trace Account Stephen J. Ceci Cornell University and Maggie
More informationImplicit Conflict Detection During Decision Making
Implicit Conflict Detection During Decision Making Wim De Neys (Wim.Deneys@psy.kuleuven.be) Department of Psychology, K.U.Leuven, Tiensestraat 102 3000 Leuven, Belgium Abstract Popular dual process theories
More informationChance. May 11, Chance Behavior The Idea of Probability Myths About Chance Behavior The Real Law of Averages Personal Probabilities
Chance May 11, 2012 Chance Behavior The Idea of Probability Myths About Chance Behavior The Real Law of Averages Personal Probabilities 1.0 Chance Behavior 16 pre-verbal infants separately watch a puppet
More informationDRAFT (Final) Concept Paper On choosing appropriate estimands and defining sensitivity analyses in confirmatory clinical trials
DRAFT (Final) Concept Paper On choosing appropriate estimands and defining sensitivity analyses in confirmatory clinical trials EFSPI Comments Page General Priority (H/M/L) Comment The concept to develop
More informationInsight Assessment Measuring Thinking Worldwide
California Critical Thinking Skills Test (CCTST). The CCTST measures the reasoning skills human beings use in the process of reflectively deciding what to believe or what to do. Skill/Attribute Name SE
More informationAnswer the questions on the handout labeled: Four Famous Reasoning Problems. Try not to remember what you may have read about these problems!
Classroom Experiment Answer the questions on the handout labeled: Four Famous Reasoning Problems Try not to remember what you may have read about these problems! Psych 466, Miyamoto, Aut '17 1 The Representativeness
More informationVIRGINIA LAW REVIEW IN BRIEF
VIRGINIA LAW REVIEW IN BRIEF VOLUME 96 JUNE 15, 2010 PAGES 35 39 REPLY GOOD INTENTIONS MATTER Katharine T. Bartlett * W HILE writing the article to which Professors Mitchell and Bielby have published responses,
More informationStudent Performance Q&A:
Student Performance Q&A: 2009 AP Statistics Free-Response Questions The following comments on the 2009 free-response questions for AP Statistics were written by the Chief Reader, Christine Franklin of
More informationWhen Falsification is the Only Path to Truth
When Falsification is the Only Path to Truth Michelle Cowley (cowleym@tcd.ie) Psychology Department, University of Dublin, Trinity College, Dublin, Ireland Ruth M.J. Byrne (rmbyrne@tcd.ie) Psychology Department,
More informationCan Bayesian models have normative pull on human reasoners?
Can Bayesian models have normative pull on human reasoners? Frank Zenker 1,2,3 1 Lund University, Department of Philosophy & Cognitive Science, LUX, Box 192, 22100 Lund, Sweden 2 Universität Konstanz,
More information2014 Philosophy. National 5. Finalised Marking Instructions
National Qualifications 2014 2014 Philosophy National 5 Finalised Marking Instructions Scottish Qualifications Authority 2014 The information in this publication may be reproduced to support SQA qualifications
More informationTechnical Whitepaper
Technical Whitepaper July, 2001 Prorating Scale Scores Consequential analysis using scales from: BDI (Beck Depression Inventory) NAS (Novaco Anger Scales) STAXI (State-Trait Anxiety Inventory) PIP (Psychotic
More informationASSIGNMENT 2. Question 4.1 In each of the following situations, describe a sample space S for the random phenomenon.
ASSIGNMENT 2 MGCR 271 SUMMER 2009 - DUE THURSDAY, MAY 21, 2009 AT 18:05 IN CLASS Question 4.1 In each of the following situations, describe a sample space S for the random phenomenon. (1) A new business
More informationCHAPTER THIRTEEN. Data Analysis and Interpretation: Part II.Tests of Statistical Significance and the Analysis Story CHAPTER OUTLINE
CHAPTER THIRTEEN Data Analysis and Interpretation: Part II.Tests of Statistical Significance and the Analysis Story CHAPTER OUTLINE OVERVIEW NULL HYPOTHESIS SIGNIFICANCE TESTING (NHST) EXPERIMENTAL SENSITIVITY
More informationChecking the counterarguments confirms that publication bias contaminated studies relating social class and unethical behavior
1 Checking the counterarguments confirms that publication bias contaminated studies relating social class and unethical behavior Gregory Francis Department of Psychological Sciences Purdue University gfrancis@purdue.edu
More informationEffect of Choice Set on Valuation of Risky Prospects
Effect of Choice Set on Valuation of Risky Prospects Neil Stewart (neil.stewart@warwick.ac.uk) Nick Chater (nick.chater@warwick.ac.uk) Henry P. Stott (hstott@owc.com) Department of Psychology, University
More informationProvided for non-commercial research and educational use only. Not for reproduction, distribution or commercial use.
Provided for non-commercial research and educational use only. Not for reproduction, distribution or commercial use. This chapter was originally published in the book Handbook of Experimental Economics
More informationQUALITY IN STRATEGIC COST ADVICE: THE EFFECT OF ANCHORING AND ADJUSTMENT
QUALITY IN STRATEGIC COST ADVICE: THE EFFECT OF ANCHORING AND ADJUSTMENT Chris Fortune 1 and Mel Lees 2 1 School of the Built Environment, Liverpool John Moores University, Clarence Street, Liverpool L3
More informationLearning probability with Fathom Óscar Chávez University of Missouri Columbia
Learning probability with Fathom Óscar Chávez University of Missouri Columbia Interface 2007 Standards, Curriculum, Assessment, Instruction: Forces Aligned to Increase Student Understanding Osage Beach,
More information